VVT Tuning Noise - ADDED: Comparison Video

[Luckrider]

Horrible video, but here is the note that everyone is talking about as observed at part throttle in 4th gear where it is most present.

[ame="https://www.youtube.com/watch?v=BwKDJyYLggo"]4000rpm Note Change - YouTube[/ame]

Edit: I do have a "better" video, but the noise isn't as clear. I'll upload that one too if anyone wants.

[vtmike]

So, are there 2 things going on here. A page back @SkullWorks mentions having the noise happen while tuning his car using only port injection, indicating its the cam phasing. @Element Tuning talks about the hp fuel pump having difficulty adjusting to rapid changes in intake air flow.

Is the heat shield rattle noise the cam phasing, and a ticking noise the hp fuel pump?

Sent from my Nexus 7 using Tapatalk

[RickyB]

Originally Posted by Dezoris

I was thinking this in the car today, the engineers and people responsible for setting up this car had to have a reason for leaving a nasty dip there. If it was so easy to tune out you would think they would have done it.

Your explanation makes the most sense at this point.

The whole design of the high pressure fuel system is a little weird and the only thing I can figure is that it was done to save costs initially and then it bit them later during testing and forced them to de-tune the midrange to prevent them from having to replace 10-20% of the fuel pumps and/or cams in the first year of service.

For those who are not familiar with the setup, there is a three lobe cam on the driver's side (US car) intake camshaft which drives the high pressure pump. This means that the system gets three pumps for every four squirts the direct injectors deliver. This requires the high pressure fuel system to have some reserve and even then it will cause strange pressure waves to constantly flow through the system. To make matters worse, the ECU must actuate the high pressure dump valve at just the right time to build proper pressure, but getting this timing exactly right requires precise knowledge of the position of the intake camshaft. As the oil activated variable cam system adjusts oil pressure to move the camshaft, the ECU must wait until the next tooth on the cam sync pickup flies under the sensor to know how fast the cam moved. Since there are only three teeth on the cam sensor, the cam must rotate up to 180 degrees (because the pattern is 90 degrees per tooth with one completely missing) before the ECU can know for sure the position of the cam. This is a full 360 degrees at the crank, so the ECU has to go blind for up to an entire crank rotation and make an educated guess as to when the dump valve needs to be activated. If the cam is not being asked to change location very fast, this is not a big deal, but if the tuner gets a bit too aggressive and commands a fast advance or retard of the cam, the dump valve timing is likely to be off enough to cause the system to lose pressure.

I expect that this design will be causing tuners and owners alike grief for a long time to come. Tuners who try to tune out the torque dip will find that their tunes work great on some cars and not others. All it takes is a marginal pump or a cam that tends to move through its range a bit faster or slower than average or perhaps even oil that is a little bit outside the right viscosity range to cause problems that result in either pump noises, fuel system noises or even real knock because the direct injection fuel pressure drops too far to deliver enough fuel to keep the engine from pinging.

I will not be surprised if the next refresh of the engine brings substantial changes to the high pressure fuel system.

[Shiv@Openflash]

Hi guys,
Here's an OFT datalog I took today running on our standard Stage 2 EL map on our shop car. I logged several WOT pulls between ~3000 and 5000prm. I did a few runs with the Phathom ESC turned off and a few runs with it turned on (and making 5psi of boost) to see if we are dealing with any inherent DI fuel system limitation. You can see the difference in Engine Load with the SC off and on. Even when running on 100% DI, at 5psi of boost, on standard NA map, there is no evidence of fuel system limitation. Fuel pressure remains steady (as one can expect) and on-target. We have been tuning direct injection BMW motors since 2006 and we have seen our fair share of tuning induced HPFP limitations. The FA20 fuel system, on the other hand, is behaving wonderfully.

I logged Fuel rail pressure, target fuel rail pressure, vvt intake, vvt exhaust, vvt intake target, vvt exhaust target, injection time PI, injection time DI, AFR and commanded AFR. Enhanced data logging FTW.

For your entertainment:

http://www.datazap.me/u/shivvishnu/v...ta=1-3-4-11-12

[Element Tuning]

RickyB, we have our FA20 engine apart (one of them) and we might be able to make a modification to the cam lobes (DI fuel pump) as we discussed. The pressure drop/fluctuations are mostly a low rpm issue but I am seeing limitations in holding high rpm pressure also where this could help.

I'll have a spare set of cams so I may attempt this and run a short durability test which is my main concern when modifying a camshaft.

Thanks,
Phil Grabow

[regal]

Originally Posted by Dezoris

I was thinking this in the car today, the engineers and people responsible for setting up this car had to have a reason for leaving a nasty dip there. If it was so easy to tune out you would think they would have done it.

Your explanation makes the most sense at this point.

Its called passing emission cert., there was a post last year with the exact test they run. EPA certification is a lot more involved than the states stickers. But I wouldn't go on ignoring the post above that this valve timing is controlled by oil pressures/selenoids, viscosity has to play a role.


I'm glad for this thread though because my 100% stock FRS has the sound comes in at 5k and sound exactly like an 1/8" hole in the exhaust manifold. And it didn't start till about 5k miles.


This is a Subaru, run it hard and you should fully expect to have take off the heads a lot sooner than a Toyota or Honda. Whether it be a screen in the AVCS, a head gasket, DI injectors, you name it. I will not be surprised if I have to pull the heads before 100k (I changed my predict from 30k)

[arghx7]

If there really is a significant interaction between transient cam phasing control and transient DI fuel pressure control, it's due to limitations in the feed forward control of the DI spill valve/dump valve/whatever you want to call it.

Originally Posted by RickyB

The whole design of the high pressure fuel system is a little weird and the only thing I can figure is that it was done to save costs initially and then it bit them later during testing and forced them to de-tune the midrange to prevent them from having to replace 10-20% of the fuel pumps and/or cams in the first year of service.

I can see how you would feel this way about the high pressure fuel pump, but I can assure you that a 3 lobe design with adjustable solenoid delivery angle is basically a standard design used all across the industry. Of course there are some variations due to intellectual property concerns.

For those who are not familiar with the setup, there is a three lobe cam on the driver's side (US car) intake camshaft which drives the high pressure pump. This means that the system gets three pumps for every four squirts the direct injectors deliver.

Normal. There are gazillions of engines running around with 3 lobe pumps, although there are a lot of 4 lobe pumps too. The highest flow DI pumps tend to have 4 lobe designs, but I suspect the 3 lobe design has less friction because the cam follower isn't climbing up and down a lob as many times. If it really is a friction issue, remember that friction kills fuel economy, torque, and power.

This requires the high pressure fuel system to have some reserve and even then it will cause strange pressure waves to constantly flow through the system. To make matters worse, the ECU must actuate the high pressure dump valve at just the right time to build proper pressure,

This is a standard design. It's called solenoid delivery angle and current control. The principle of the hardware is common among high pressure fuel pumps.



In the example above, you have three things. You have the red curve, the green box, and the shaded orange area. The shaded orange area is the portion of the delivery stroke which is being used. The green box is basically the duration of solenoid energization. The red is the current profile of the solenoid. The "width" of the green box is fixed. So, the ECU calculates how much of an orange shaded area it wants. The shaded area is based on the position of the green box. The green box is based on the current control (red dotted line). There are governing feed forward equations and calibrations, plus feedback gains. These are primarily set on a motoring rig and on an engine dyno.

The ECU has to calculate all three of these things (orange, green, red) and coordinate them with other actuators. Somebody at Subaru/Toyota/whoever said it was "good enough," and it probably was for a stock car with a stock calibration. This isn't something that can be easily reverse engineered and adjusted in the aftermarket, and I don't see Subaru ever bothering to do so.

but getting this timing exactly right requires precise knowledge of the position of the intake camshaft.

So do many other things in engine control, including spark and injection timing (because the cam position tells you which piston is on which stroke).

As the oil activated variable cam system adjusts oil pressure to move the camshaft, the ECU must wait until the next tooth on the cam sync pickup flies under the sensor to know how fast the cam moved.

This is all modeled and calibrated inside the ECU. Now I'm not saying Toyota, or Subaru, or whoever did as good of a job as we would have liked.

Since there are only three teeth on the cam sensor, the cam must rotate up to 180 degrees (because the pattern is 90 degrees per tooth with one completely missing) before the ECU can know for sure the position of the cam.

That's just for sync--the rotation and acceleration of the cam and crank are still modeled. That's required for the OBD monitors at least.

This is a full 360 degrees at the crank, so the ECU has to go blind for up to an entire crank rotation and make an educated guess as to when the dump valve needs to be activated. If the cam is not being asked to change location very fast, this is not a big deal, but if the tuner gets a bit too aggressive and commands a fast advance or retard of the cam, the dump valve timing is likely to be off enough to cause the system to lose pressure.

Just about every Hall-type cam position sensor is that way. These fuel pump controls are based on a feed forward calculation that usually comes from the supplier and can be adjusted in-house if desired by the OEM. I wouldn't be surprised if they just dropped in the supplier's control system and did a minimal amount of validation on an engine dyno (step-response tests) then judge it good enough. See above.

I expect that this design will be causing tuners and owners alike grief for a long time to come. Tuners who try to tune out the torque dip will find that their tunes work great on some cars and not others. All it takes is a marginal pump or a cam that tends to move through its range a bit faster or slower than average or perhaps even oil that is a little bit outside the right viscosity range to cause problems that result in either pump noises, fuel system noises or even real knock because the direct injection fuel pressure drops too far to deliver enough fuel to keep the engine from pinging.

This might happen, this might not. It's probably not the hardware's fault except for some of the parts that have had known issues. There are all sorts of fuel system controls in the ECU that nobody has reverse engineered yet as I've alluded to above.

I will not be surprised if the next refresh of the engine brings substantial changes to the high pressure fuel system.

While we can hope for a more robust hardware design or something better on the software side, Subaru or Toyota is not going to devote any resources to this unless there are some serious warranty claims.

[RickyB]

Originally Posted by arghx7

If there really is a significant interaction between transient cam phasing control and transient DI fuel pressure control, it's due to limitations in the feed forward control of the DI spill valve/dump valve/whatever you want to call it.

Yes, I am quite familiar with these systems and wrote the firmware in the Hydra to handle the control of the fuel pump pressure. While I know that there are probably some things that I can do to improve the predictability of the algorithm slightly, any engine management system looking at the position of a particular shaft (crank or cam) is limited by how many teeth are on the sensor of the shaft to update the actual position vs. best guess estimate of said position.

The spark and direct injection timing are based on the position of the crank provided by the trigger wheel which has 36-2-1 teeth, so it gets an update roughly every 10 degrees of crank revolution. The gaps on the wheel are carefully placed so that nothing important has to happen during the gap. This makes the job of firing the injectors and igniters at the proper times "easy" for the ECU.

Subaru/Toyota in this case mixed a 3 lobe system (120-120-120 degree events) with the STI 4-1 cam sensor (90-90-180 events). This mean that any way you try to line things up, you are going to end up with one spill valve event taking place a relatively long time after the last sync up. During an aggressive cam advance attempt, the actual cam position (which depends on many chaotic factors requiring constant closed loop control) may be many degrees off from the ECU's predicted cam position, so that the green pulse will not overlap with as much of the orange ramp as the ECU expects. In fact, the green pulse may happen a little too soon during the intake stroke and cause the valve to shut off flow into the plunger, which can cause the rocker to stop its travel and then get slammed by the cam on its way back up, causing all sorts of "marble hitting a can" type noises.

So while the system can be made to "work," and probably will do great in an 80whp econo-box, my theory is that Subaru/Toyota had issues with it tracking well enough during fast cam angle changes and ended up de-tuning the midrange (hence the dreaded dip) by not advancing the cam as fast as the motor needs to make maximum torque. I could be wrong about this, but I stared at quite a few scopes from several dyno pulls we did with the factory ECU while reverse engineering the high pressure fuel pump control and the scopes definitely show that the valve control points are off quite a bit more during the intake cam tooth gap than they are during the rest of the cycle.

The manufacturer can address these issues by either going to a four lobe pump, which has the downside of increasing friction and reducing economy as you state, or changing the intake cam sync wheel to a different pattern that allows the ECU to get a proper reading of the camshaft location at a point that is always just before the green pulse has to start on all three of the dump valve events. Another approach is to drive the pump off the crankshaft or the front portion of the cam before the VVT actuator whose position can be deduced from the crank trigger. Yet another approach would be to install two fuel pumps each driven by a two lobe cam on each side of the engine (go with separate left and right bank high pressure fuel rails).

[arghx7]

Originally Posted by RickyB

Yes, I am quite familiar with these systems and wrote the firmware in the Hydra to handle the control of the fuel pump pressure. While I know that there are probably some things that I can do to improve the predictability of the algorithm slightly, any engine management system looking at the position of a particular shaft (crank or cam) is limited by how many teeth are on the sensor of the shaft to update the actual position vs. best guess estimate of said position.

The spark and direct injection timing are based on the position of the crank provided by the trigger wheel which has 36-2-1 teeth, so it gets an update roughly every 10 degrees of crank revolution. The gaps on the wheel are carefully placed so that nothing important has to happen during the gap. This makes the job of firing the injectors and igniters at the proper times "easy" for the ECU.

Agreed. 36-2-1 , or 57 teeth, or all the other arrangements for modern crank position sensors, is still better than a Hall sensor with a few slots in it.

Subaru/Toyota in this case mixed a 3 lobe system (120-120-120 degree events) with the STI 4-1 cam sensor (90-90-180 events). This mean that any way you try to line things up, you are going to end up with one spill valve event taking place a relatively long time after the last sync up.

Yup, that's an inherent hardware issue.

During an aggressive cam advance attempt, the actual cam position (which depends on many chaotic factors requiring constant closed loop control) may be many degrees off from the ECU's predicted cam position, so that the green pulse will not overlap with as much of the orange ramp as the ECU expects.

Yup. And the problem is the predicted cam position and its relationship to the control of the green pulse.

In fact, the green pulse may happen a little too soon during the intake stroke and cause the valve to shut off flow into the plunger, which can cause the rocker to stop its travel and then get slammed by the cam on its way back up, causing all sorts of "marble hitting a can" type noises.

Yup, agreed this could happen.

So while the system can be made to "work," and probably will do great in an 80whp econo-box, my theory is that Subaru/Toyota had issues with it tracking well enough during fast cam angle changes

I wouldn't be surprised if Subaru/Toyota spent very little time testing fast cam angle changes, because the vast majority of fuel pressure and cam phasing testing is done in a steady state at the early stage of an engine development program. We're used to quick transient WOT pulls on a dynojet, but in a lab that's done way later after much of the hardware and software parameters have been frozen. Then they probably handed a pile of turd (metaphorically speaking) to the NVH guys and said "here, make this sound not too obnoxious, and don't spend much money doing it"

Meanwhile, most fuel system suppliers aren't going to give any software or calibration help to an OEM beyond what is exactly specified in the purchase order. Considering most of the controls appear to be Subaru/Denso based, and Subaru has only put DI into production recently, you can't expect them to have had much of an understanding of DI fuel pump control. So they are going to lean on the supplier, and the supplier isn't going to spend any more resources than they are contractually obligate to.

and ended up de-tuning the midrange (hence the dreaded dip) by not advancing the cam as fast as the motor needs to make maximum torque.

That's an interesting theory, but I'd say I'm pretty skeptical, especially considering almost every n/a DI engine in production today has some kind of torque dip.

These days the vast majority of cam phasing maps, fuel pressure, and injection timing maps are set in the steady state using automated test schedules with model-based Design of Experiments methodology. There could have been engine-out emissions, combustion stability, or other concerns found in steady state mapping that resulted in the cam phasing we see in the stock calibration. That's in addition to all the other factors that have been discussed to death on this forum regarding the torque dip.

I could be wrong about this, but I stared at quite a few scopes from several dyno pulls we did with the factory ECU while reverse engineering the high pressure fuel pump control and the scopes definitely show that the valve control points are off quite a bit more during the intake cam tooth gap than they are during the rest of the cycle.

That wouldn't surprise me either, but it's hard to say if their control capability is any better or worse than other engines out there.

The manufacturer can address these issues by either going to a four lobe pump, which has the downside of increasing friction and reducing economy as you state, or changing the intake cam sync wheel to a different pattern that allows the ECU to get a proper reading of the camshaft location at a point that is always just before the green pulse has to start on all three of the dump valve events. Another approach is to drive the pump off the crankshaft or the front portion of the cam before the VVT actuator whose position can be deduced from the crank trigger. Yet another approach would be to install two fuel pumps each driven by a two lobe cam on each side of the engine (go with separate left and right bank high pressure fuel rails).

I still believe that better feed forward/model-based control can improve the situation, especially when you have such complicated systems out there like Valvetronic and MultiAir, but this is becoming a mostly academic discussion anyway.

The people who design the software and controls blame the hardware if it doesn't meet targets ("this is a dumb design").

The people who design the hardware blame the software and calibration if it doesn't meet targets ("you're doing it wrong").

[RickyB]

Originally Posted by arghx7

The people who design the software and controls blame the hardware if it doesn't meet targets ("this is a dumb design").

The people who design the hardware blame the software and calibration if it doesn't meet targets ("you're doing it wrong").

Yup. Been there, done that on both sides of the fence

You are right that there is little to be gained by speculating how and why certain things were done. The important thing is for tuners/owners/builders to be aware of how the system works and understand its limitations so that they don't get "surprised" when they try to push towards or even beyond the capabilities of the hardware/software.

[Element Tuning]

@arghx7

I always respect your input. Because it's normal or the industry standard doesn't mean it will work how we need it to in the aftermarket tuning world as we expect more than what's given to us. Shiv's logs are a little slow on the data rate but you can see the fuel pressure drop every time there is request for a significant jump in intake advance. So far from the cars I've tested this pressure dip is normal in the 2500-3500 RPM area. I'm really not sure what the OEM reflash tuners have access to in terms of fine tuning the DI pump and Spill valve compared to the Hydra EMS I'm using which has a lot of tuning ability for the pump.

This is only a real issue at lower RPM. The pressure wave is not normal however and I'm sure you're aware at how much noise that will make. Most of the people in this thread just have normal intake noise but those who've mentioned "ticking" or "marbles" it's related to the DI pressure waves. If a 4 lobe cam would help stabilize the pressure dip at 3000 RPM and maybe hold better at high RPM that would only benefit performance as I can get more torque down low. On my base maps I have to detune this area some to deal with differences in peoples cars but some have been able to squeak out a little more DI pressure there and it improves torque in that range.

Thanks,
Phil Grabow

[arghx7]

Yeah it's unfortunate that we don't have a nice broad selection of bolt-in aftermarket high pressure fuel pumps to play around with. It'll probably be a while before that happens. Direct injection is probably where conventional fuel injection was 25 years ago in terms of part availability.

[Dezoris]

So can you guys sum up the solution to the marbling for tuners? I understand some may not have it while others do, like in my case.

Is it just a matter of de-tuning the cam angle change to less degrees?

[arghx7]

Originally Posted by Dezoris

So can you guys sum up the solution to the marbling for tuners? I understand some may not have it while others do, like in my case.

Is it just a matter of de-tuning the cam angle change to less degrees?

I think there's still a debate (well, in my mind at least) whether the noise is a problem for everyone to be concerned about, beyond being annoying. Is it like a creaking suspension--not necessarily a sign of anything actually wrong, just undesireable, or is it like spark knock or rod knock, indicating some damage could be occurring?